Building Confidence in Long-Term Safety Assessments Using Complimentary Computer Codes

Abstract

Canadian Regulatory Documents identify the need to have confidence in long-term assessment models. One of the activities identified to enhance confidence in the long-term assessments is performing independent predictions using entirely different computing tools. Performing assessments of the same scenario using the different computer codes, such as AMBER¹ and RESRAD-OFFSITE², can provide enhanced confidence for the long-term safety assessment.

The AMBER code is a flexible software tool that allows users to implement their own mathematical models to represent contaminant transport within a compartment model approach to estimate radiological and non-radiological contaminant fluxes and concentrations for the environmental media within the various pathways as well as the radiological consequences and the peak time of occurrence.

RESRAD-OFFSITE is a comprehensive software package for estimating radiological soil cleanup values (radionuclide activity concentrations) and the radiological dose consequences to a human receptor located directly above the primary contamination (onsite) or located in the vicinity of the primary contamination (offsite).

A long-term safety assessment for the in-situ disposal of the Nuclear Power Demonstration (NPD) reactor was prepared using AMBER modelled the radiological effects to the public for multiple scenarios for a disposal of a facility. A second assessment was performed using RESRAD-OFFSITE on two of the AMBER models to provide enhanced confidence in the results.

The results of the AMBER and RESRAD-OFFSITE were compared to identify variations in the results. Reviewing the variations identified how differences in the modelling influenced the long-term dose consequences. Overall, comparison between the AMBER and RESRAD-OFFSITE results/output can enhance the confidence in the long-term modeling by showing how two different computer codes using different calculation methods produce similar results given the modelling assumptions and code differences.